Each row states what the design is, when it fits, how collection is structured, what evidence comes out, what analysis is available, and how it is governed — followed by the routes an answer can arrive on.
Structured survey
Live
Also called Household survey, Population survey
A defined questionnaire put to a defined sample, so results can be reported as numbers with a stated confidence and a stated method.
- When to use it
- When you need to measure something across a population and defend the figure — coverage, access, prevalence, satisfaction, or a results-framework indicator.
- How collection is structured
- A study carries its instrument and question bank, a sampling plan with strata, and a weight set. Responses arrive against that instrument from any live channel.
- Evidence it produces
- Weighted and unweighted estimates, confidence intervals, disaggregation, and a methodology statement that reports what the study did not declare rather than omitting the line.
- Analysis available
- Descriptives, proportions and means with Wilson and Welch–Satterthwaite intervals, and significance tests. Every figure is computed by deterministic code, never by a language model.
- Governance and quality
- A non-probability sample is refused representativeness regardless of what the study declares, and cannot be post-stratified. Analysis runs pass quality gates and carry approvals.
- Typical outputs
- Statistical report · Indicator dashboard · Dataset and codebook · Baseline or endline publication
- What “Live” means
- VoiceInsights runs this design end to end on governed structures built for it, with its own interfaces.
Phone CallWhatsAppSMSWeb linkOffline appRoutes an answer can arrive on. A channel carries the design; it is not part of it.
Facility assessment
Live
Also called Service delivery assessment, Site assessment
A structured assessment of a service point — a clinic, a school, a water point, an office — carried out at the site by a trained assessor.
- When to use it
- When the unit you are measuring is a facility rather than a person: readiness, staffing, stock, infrastructure, or compliance with a standard.
- How collection is structured
- The same study, instrument and sampling structures as a structured survey, with the facility as the unit of analysis and assessment instruments drawn from the sector libraries.
- Evidence it produces
- Scored assessments per site, comparison across sites and administrative areas, and the geographic record of where each assessment was taken.
- Analysis available
- The same deterministic statistical engine, with results reported by site, district and country.
- Governance and quality
- Location is treated as sensitive: precise coordinates require an explicit export permission, separate from the permission to read the assessment.
- Typical outputs
- Facility assessment report · Site comparison dashboard · Readiness scorecard
- What “Live” means
- VoiceInsights runs this design end to end on governed structures built for it, with its own interfaces.
Web linkOffline appRoutes an answer can arrive on. A channel carries the design; it is not part of it.
Citizen feedback
Live
Also called Community feedback, Service user feedback
An open route for the people a service exists for to say what is happening, in their own language, on the device they already have.
What you should know
- Results describe the people who chose to respond. They are not generalisable to a population, and the platform refuses to weight them as though they were.
- When to use it
- When you need continuous signal rather than a periodic study — accountability lines, grievance channels, service ratings, or community scorecards.
- How collection is structured
- Runs on the autonomous channels: a phone call, a WhatsApp message or voice note, an SMS, or a shared web link. There is no sampling frame and the platform does not invent one.
- Evidence it produces
- Responses with their channel, declared language and consent record, and quotations that remain attached to the response they came from.
- Analysis available
- Thematic analysis through the qualitative model, and counts by channel, area and period. No representativeness is claimed from a self-selected respondent set.
- Governance and quality
- A citizen-feedback study has no sampling frame, and the completeness projection reports that as NOT_APPLICABLE rather than as a gap — so it is never scored down for lacking something the design does not have.
- Typical outputs
- Community scorecard · Feedback dashboard · Thematic brief · Accountability report
- What “Live” means
- VoiceInsights runs this design end to end on governed structures built for it, with its own interfaces.
Phone CallWhatsAppSMSWeb linkRoutes an answer can arrive on. A channel carries the design; it is not part of it.
Qualitative research
Live
Also called Interview and discussion research
Structured analysis of what people actually said — coded against a codebook, organised into themes, and traceable back to the words that produced them.
- When to use it
- When you need to understand why something is happening, not only how often: barriers, decisions, experience, and the language people use for it.
- How collection is structured
- Recordings and transcripts arrive as media uploads or through the field workspace. Interview formats — key informant, in-depth and focus group — are formats within this mode, not modes of their own.
- Evidence it produces
- A versioned codebook, codes applied to passages, themes assembled from those codes, and quotations that keep their link to the original response.
- Analysis available
- Codebook development, coding, theme assembly and a qualitative summary. Interpretation stays attributable to the passage it rests on.
- Governance and quality
- A quotation is never separated from its source, and a translation is never presented as the original. Where a translated quotation is published, the original wording is published beside it.
- Typical outputs
- Thematic report · Evidence pack with quotations · Qualitative annex · Knowledge product
- Formats it carries
- Key informant interview (KII), In-depth interview (IDI), Focus group discussion (FGD). These are ways of gathering material within this mode, not modes of their own.
- What “Live” means
- VoiceInsights runs this design end to end on governed structures built for it, with its own interfaces.
Phone CallWhatsAppOffline appRoutes an answer can arrive on. A channel carries the design; it is not part of it.
Mixed methods
Supported
Also called Combined quantitative and qualitative design
One study that carries both strands — a measured quantitative component and a coded qualitative component — held together by a single study record.
What you should know
- There is no automated triangulation or joint-display engine. Bringing the two strands into one narrative is analyst work, supported by the platform rather than performed by it.
- The mixed-method synthesis tables that once existed belong to a deprecated schema that no running code reads. They are not a capability and are not presented as one.
- When to use it
- When a number alone will not carry the decision and an account alone will not carry the scale: an evaluation that must both measure change and explain it.
- How collection is structured
- A single study declares the mixed-method paradigm and carries its instrument and sample for the quantitative strand alongside its codebook and transcripts for the qualitative strand.
- Evidence it produces
- Both strands, on one study, retaining their own methods and their own limits. Statistical results stay statistical; coded passages stay attributable.
- Analysis available
- Each strand is analysed by its own engine — the deterministic statistics engine and the qualitative coding model — against the same study.
- Governance and quality
- The two strands are never merged into a single score. A qualitative theme does not become a statistic because it appears in the same study as one.
- Typical outputs
- Mixed-methods evaluation report · Evidence pack combining figures and quotations
- What “Supported” means
- VoiceInsights runs this design using the platform’s shared research structures. There is no workspace built specifically for it, so some steps are configured with your implementation team rather than driven from a dedicated screen.
Phone CallWhatsAppSMSWeb linkOffline appRoutes an answer can arrive on. A channel carries the design; it is not part of it.
Observation and field assessment
Supported
Also called Direct observation, Field verification
Recording what a trained field worker sees rather than what a respondent reports — verification, spot checks, and condition assessments captured on site.
What you should know
- There is no observation-specific instrument type. An observation checklist is built as a study instrument, which works but means the platform does not distinguish an observed value from a reported one in the data model.
- When to use it
- When self-reported answers are not sufficient evidence, or when the thing being assessed is physical: infrastructure condition, stock on a shelf, a practice observed.
- How collection is structured
- Captured through the offline field workspace, which records the observation against its assignment together with location, photographs, audio and the consent record, and syncs when a connection returns.
- Evidence it produces
- Observations bound to a place and a time, with the media captured alongside them and a sync ledger showing when each record arrived.
- Analysis available
- Observations analyse as structured responses. Geographic aggregation is available to administrative area.
- Governance and quality
- Each observation carries the assignment and field worker it came from, so a record can always be traced to who took it and where.
- Typical outputs
- Field assessment report · Verification dashboard · Geographic summary
- What “Supported” means
- VoiceInsights runs this design using the platform’s shared research structures. There is no workspace built specifically for it, so some steps are configured with your implementation team rather than driven from a dedicated screen.
Offline appRoutes an answer can arrive on. A channel carries the design; it is not part of it.
Longitudinal and panel
Supported
Also called Repeat waves, Cohort follow-up
The same study run again over time, with the waves recorded as waves and the same respondents linked across them where the design follows individuals.
What you should know
- Waves and attrition are available through the research API and the data model. There is no panel-management workspace, so wave setup and cohort tracking are operated with your implementation team rather than from a dedicated screen.
- When to use it
- When change over time in the same units is the finding — a cohort followed through a programme, or a measure repeated on a schedule.
- How collection is structured
- A study declares its waves; each wave carries its own fieldwork. Panel links hold the relationship between a unit in one wave and the same unit in the next.
- Evidence it produces
- Wave-by-wave results, the panel linkage, and an attrition assessment showing who was lost between waves.
- Analysis available
- Each wave analyses through the same statistical engine. Attrition is computed rather than assumed, so loss between waves is visible instead of silently reducing the base.
- Governance and quality
- A study declaring a panel design cannot also declare a non-panel temporality; the contradiction is refused at validation rather than carried into analysis.
- Typical outputs
- Wave comparison report · Cohort analysis · Attrition annex
- What “Supported” means
- VoiceInsights runs this design using the platform’s shared research structures. There is no workspace built specifically for it, so some steps are configured with your implementation team rather than driven from a dedicated screen.
Phone CallWhatsAppSMSWeb linkOffline appRoutes an answer can arrive on. A channel carries the design; it is not part of it.